A scene may be reconstructed using image data captured by an image capture device at various positions within the scene. The image capture device may be positioned throughout the scene in order to capture the scene from many diverse perspectives and vantage points. Ideally, the scene would be observed from all possible angles in order to generate the best-possible recreation, but doing so is time-consuming and resource-intensive. Conversely, capturing data from too few positions creates difficulties when reconstructing the scene, as it becomes difficult to identify image features and anchor points between multiple images if the image data captured at various positions lacks commonality.
It is with respect to these and other general considerations that the aspects disclosed herein have been made. Also, although relatively specific problems may be discussed, it should be understood that the examples should not be limited to solving the specific problems identified in the background or elsewhere in this disclosure.